2024, Vol. 7, Special Issue 4

author(s)
Sellaperumal Pazhanivelan, G Srinivasan, NS Sudarmanian, S Thirumeninathan, Kancheti Mrunalini, Ragunath Kaliyaperumal, Satheesh Sakthivel and Manoj Kumar Yadav

abstract
Cotton is cultivated in a discrete pattern and the area estimation using traditional methods are laborious and time consuming. Further, cotton cultivation is restricted to monsoonal period (September-October) which coincides with dense clouds causing diffused imageries in optical remote sensing. Such problems are collectively addressed using Synthetic Aperture Radar (SAR) which is capable of penetrating the cloud cover providing more realistic data for area assessment. With this in view, a study was undertaken in southern peninsular India in the State of Tamil Nadu where cotton is primarily cultivated in three districts (Ariyalur, Cuddalore, Perambalur). The Sentinel 1A satellite imageries were extracted periodically throughout the cropping season from the point of sowing till the harvest at an interval of 12 days (August 2019 to January 2020) and processed using MAPscape-RICE software. The satellite data were validated using 200 ground truth points spread across the study area besides twenty-five monitoring sites for accurate assessment. Multi-Temporal features viz., max, min, mean, max date, min date and span ratio were extracted from VH polarization of Sentinel 1A GRD data to classify cotton pixels in the study area using parameterized classification approach. Overall data suggest that the SAR predicts the cotton area classification accuracy to the tune of 89.0 per cent with a reliability of 88.1 per cent and a kappa score of 0.78. Consequently, the classified cotton area of Ariyalur, Cuddalore and Perambalur districts were estimated as 5371, 4990 and 14,162 ha, respectively. The observed cotton yield ranged from 1308 to 2340 kg ha-1 and simulated cotton yield ranged from 1147 to 2799 kg ha-1 with RMSE and NRMSE of 237.8 kg ha-1 and 12.24 per cent, respectively with the overall agreement of 87.76 per cent. The remote sensing-based spatial yield ranged from 1080 to 2745 kg ha-1 across the study area with a mean cotton yield of 2070 kg ha-1. This study clearly demonstrates that SAR can be effectively employed for an accurate measurement of cotton cropped area and yield estimation with less manpower and time besides more precision.